Descriptive, Substantive, and Symbolic Effects of Gender Quotas in Latin American National Elections
Andreas Kotsadama
Måns Nermanb
University of Oslo
University of Gothenburg
Abstract This study investigates the effects of gender quotas in national elections on political participation and public policy in Latin America. Using fixed effects we are able to confirm the findings from previous research that women in politics are correlated with these outcomes in our sample. However, we find no substantial effects of legal gender quotas in Latin America beyond mere representation. The mechanisms for these findings are scrutinized, and our findings are consistent with quota women being marginalized from power in Latin American parliaments. Keywords: Gender quotas, Latin America, Women in Parliament
a
Corresponding author. Department of Economics, University of Oslo, Box 1095 Blindern, 0317 Oslo, Norway. Email:
[email protected] b Department of Economics, University of Gothenburg, P.O. Box 640, SE 405 30, Gothenburg, Sweden. Phone: +46 (0)31-786 4720. Email:
[email protected] Acknowledgements: The paper has benefited from comments by seminar participants at the 2012 CSAE conference on Economic Development in Africa, and at a DEW seminar, University of Gothenburg. We thank Maria Escobar-Lemmon for generously sharing data on female ministers in Latin America. We would also like to thank Niklas Bengtsson, Arne Bigsten, Jon-Ivar Elstad, Henning Finseraas, Niklas Jakobsson, Andy McKay, Magne Mogstad, Måns Söderbom, and Pär Zetterberg for useful comments.
1
1
Introduction
Women account for half of the world’s population, but hold just short of one-fifth of the parliamentarian seats worldwide. Fifteen years ago the same figure stood at just eleven percent (IPU statistical archive). Academic debates about both the reasons and consequences of this change, as well as the still low share, are frequent in sociology, economics, and political science (Del Campo 2005; Fallon et al. 2012). Over these years, various affirmative actions have been introduced in more than 100 countries around the world in order to increase the number of women in politics, either by governments or by political parties of their own accord. When looking at what determines the levels of female representation today, gender quotas stand out as the factor with the most explanatory power (Tripp and Kang 2008). In terms of legal electoral quotas, Latin America has been in the forefront; in 1991 Argentina was the first democratic country to adopt a gender quota in national elections, and throughout that decade many Latin American countries followed suit (Jones 2009). As of today, eleven of the eighteen countries in Latin America have introduced gender quotas in national elections, making Latin America the region in the world with the highest share of quota countries. The arguments usually put forward for quotas are based on justice, women’s experiences, women’s interests, and the importance of female politicians as role models (Dahlerup 2003). In particular, it is argued that women have different experiences than men, or that men and women inhibit different preferences due to having conflicting interests, which has largely been confirmed by previous empirical research (e.g., Phillips 1995; Lott and Kenny 1999; Edlund and Pande 2002; Edlund et al. 2005; Miller 2008; Piscopo 2011; Finseraas et al. 2012). If women’s preferences are different than men’s, it is also possible that they may not be fully taken account of without proper female representation. Empirical evidence seems to support this too; results from cross-sectional comparisons show that increased political
2
representation of women is correlated with different spending priorities (e.g., Thomas 1991; Besley and Case 2003; Kittilson 2008; Childs and Krook 2009: Swiss et al. 2012). These results rarely imply causation though; one obvious concern is that women may be better represented in areas where voters have specific political preferences. However, there are a few studies that have used methods that help identify causal effects of women in politics on policy outcomes. Clots-Figueras (2011) studies close elections between women and men in India and finds that elected women invest more in education and Svaleryd (2009) uses longitudinal data and finds that Swedish municipal boards with increased female representation tend to increase spending on child care relative to spending on the elderly. While these investigations into the causal effects of increased female participation in elite politics are important, it is far from obvious that increasing the share of women by means of quotas would produce similar effects. For instance, quota women may feel a responsibility to represent women’s interest more strongly, or conversely work more actively against women’s issues in order not to be seen as different from non-quota politicians (Franceschet et al. 2012a). Women elected via quotas may also be more or less qualified politicians, more or less tied to the party elite, and be more or less marginalized than other female politicians (Dahlerup 2008; Franceschet and Piscopo 2008; Zetterberg 2009; Franceschet et al. 2012a) . A few studies have looked at the effects of political gender quota introductions on policies and gender norms. Of special interest for this study, Zetterberg (2009) compares the Latin American countries that had introduced quotas by 2005 to those that had not. Using a range of different measures, he finds no differences in either women’s political attitudes or their political participation. However, a limitation of his study is that the results rely on variation across countries, whereas we will also make use of variation within countries over time, allowing us to robustly control for time invariant differences across countries. With respect to policy outcomes, Chattopadhay and Duflo (2004) use the randomized introduction
3
of a gender quota at a local level in India, and find that policies on issues closer to the preferences of women were implemented more often in villages with female chiefs. Beaman et al. (2009), using the same setting as Chattopadhay and Duflo (2004), also find changing gender norms following the quota. Hence, there are claims of various effects of both women’s representation and of gender quotas. However, given that many gender norms and roles are highly context specific, one can hardly expect such findings to be the same in all societies. Furthermore, it is likely that quotas at different levels of governance also produce differing results, as decision making procedures, policy responsibilities, and closeness to the electorate differ greatly across political contexts. In the present article, we aim to identify the impacts of the extensive introductions of quotas for women in national elections in Latin American countries. Specifically, we look at the impacts on a group of outcomes that previous research has suggested that women in politics should affect, including women’s political participation and policies on e.g. education and health. By using national-level rather than local-level quotas this study will broaden the knowledge of the effects of quotas by looking at their effects in contexts typically dominated by party elites. Furthermore, the social context in the Latin American countries, with their often similar historical and socio-economic backgrounds, differs from that in other countries previously investigated. Given the potential impact and political sensitivity of the introduction of quotas in these countries, studying their effects is of great importance. Our results show that while the quotas substantially increased the number of women in parliament, we are unable to find any measurable effects on policy or political participation. However, we find that estimations using variation in the share of women in parliament not caused by quota measures often show correlations between female representation in parliament and the mentioned outcomes. This indicates either that women elected through
4
quotas to some extent differ from other elected women, or that the correlations between women in parliament and our outcomes are spurious. In either case, it suggests that this far the effects of the quotas on these issues have been at best highly inconsistent in Latin America. Further analysis also shows that we are unable to find that the quotas increased the share of women in ministerial positions, suggesting that quota parliamentarians may be largely marginalized. 2
Theory and Expectations
This section presents theory on how quotas may be effective in changing policy, how and why quotas were introduced in Latin America, and to what extent one can consider them exogenous to other developments. Based on these discussions and the existing literature, we then discuss what effects of quotas we can expect to find in the Latin American context. 2.1
Theory on Quota Effects
The motivations for implementing gender quotas point to several mechanisms through which quotas are thought to have effects beyond representation. Dahlerup (2003) presents four arguments: a justice argument, implying that women have the right to half of the representative seats as they constitute half of the population; an experience argument, implying that the gender-specific experiences of women, whether biological or socially constructed, need to be represented; an interest group argument, arguing that men and women may have conflicting interests and, consequently, men cannot represent women; and a role model argument, where the existence of female politicians is thought to help other women engage in politics. Araújo and García (2006) add to these arguments that a higher female representation may also lend further legitimacy to the democratic system and its institutions. While the justice argument is more concerned with representation per se, having different experiences or interests implies that men and women may have different preferences over policy. Since it is not obvious that women have shared interests different from those of men,
5
the experience and interest group arguments have long been debated (Sapiro 1981; Diamond and Hartsock, 1981). One particularly salient perspective is that women have similar experiences that give them different perspectives on many issues (Phillips 1995; Barnes 2012), which means that women need not have the exact same perspectives yet still share a common set of ideas that they do not share to the same extent with men (Piscopo 2011). There is ample evidence of this, and of special interest to this study Barnes (2012) finds that women in 18 provincial legislative chambers in Argentina have preferences more similar to other women than to men. However, even if men and women have different political preferences on average, and quotas lead to more women whom have political power, it is still possible that the elected quota women have preferences more similar to male than to female incumbent politicians. For instance, even with gender quotas in place it is still the political parties that nominate the candidates, which may result in women close to the incumbent (male) elites being chosen to fill out the quotas. In line with this, Franceschet and Piscopo (2008) find that the perception of many politicians in Argentina is that women elected by quotas have been placed there because they are loyal to the party or tied to a male politician. Similarly, in a later study Franceschet and Piscopo (2012) investigate the differences between quota women and non-quota men in Argentina and show that the women are of similar age and have equally high education, but have fewer children, are less likely to be married, and have less political experience. But whether the perceptions that quota women are different are correct or not, such perceptions in themselves may reduce the power or willingness of women in parliament to engage in matters concerning what is viewed as women’s issues. On the other hand, it may also be that women elected via gender quotas feel a mandate or obligation to act on behalf of women, hence conversely more strongly support such issues (Franceschet and Piscopo 2008). In other words, it seems that having more quota women in parliament may lead to more or less woman
6
friendly politicians depending on circumstances, which points to the necessity of studying their effects in different cultural and institutional contexts. However, even if we assume that there are differences in preferences for the elected quota women and the incumbent elite, the identity of the policymaker need not affect policy if the quota women do not have sufficient power or if they are more concerned with re-election or advancing their own career’s than with changing policy. That is, female legislators may not represent constituents differently than their male colleagues, as they are all trying to please the same people (Barnes 2012). This is maybe most famously formulated in Downs’ (1957) median voter theorem, which predicts that as long as candidates can commit to policies, political competition leads to an allocation that is preferred by the median voter irrespective of the identity of the politician. There is nevertheless ample evidence that the identity of the politician can indeed have an effect on policy: Pande (2003) and Besley et al. (2004) find that political reservations in local governments for disadvantaged castes in India affected the provision of public goods, and Chattopadhyay and Duflo (2004) and Clots-Figueras (2011) present similar findings for reservations for women. These effects can be better explained in alternative models such as the citizen candidate model, where political candidates cannot completely commit to a policy platform, and if they get elected they try to implement their own political agenda (e.g., Besley and Coate 1997; Chattopadhyay and Duflo 2004). The other main reason that female legislators need not affect policy even if men and women have different political preferences is that they may not have sufficient power to do so. Related to this, previous research has argued that the introductions of quotas were intended to break a male-dominated structure, and that it would be likely that the incumbent elite reacts to this (Childs and Krook 2006; Dahlerup 2006; Grey 2006; Zetterberg 2008). One proposed strategy for the maintenance of status quo would then be to marginalize women into positions with less power. In line with this, Miguel (2012) finds that Brazilian quota women
7
remain absent from positions of political influence and that they are less successful in passing legislative initiatives than are men. It is not clear, however, that such attempts would merely be reactions to increasing women in parliaments via quotas. For instance, investigating the effects of quotas on women’s political power in Mexican state legislatures, Zetterberg (2008) does not find women elected in a quota state to face greater obstacles than women elected in a no-quota state. Instead, he finds that all elected women face severe constraints. Taking this argument one step further, Fréchette et al. (2008) argue that the male incumbency advantage may even increase following gender quotas, if there is a bias among the voters for male candidates. This is so since the incumbent men will not have to compete with other men to the same extent after a quota introduction. In line with this, Heath et al. (2005) find a negative correlation between the share of elected women and the share of women in high status committees in six Latin American countries. Similarly, Meier (2012) argues that male politicians in Belgium reacted negatively to the introduction of gender quotas within political parties and therefore worked against a stronger focus on gender equality in party documents. Finally, turning to the role model argument that female politicians may spur increased political engagement among women in general, it assumes that the increase of symbolic representation for women in an elite will cause women’s political participation to rise also among the general public (e.g. Barnes and Burchard 2012). There are multiple possible mechanisms through which this may happen. It has been argued that female descriptive representation may signal that politics is not only a “man’s game” (Verba, Burns, & Schlozman, 1997) and thereby increase female political engagement (Krook, 2006). Others have argued that elite participation may affect women’s perceived legitimacy of elected bodies (Childs 2004; Schwindt-Bayer and Mishler 2005; Franceschet et al. 2012) and Schwindt-Bayer and Mishler (2005) find that citizens think that government is more democratic when women are present. Supporting this, Barnes and Burchard (2012) find that
8
that an increased representation of women in parliaments is related to mass female political participation across 20 countries in Sub-Saharan Africa, and Desposato and Norrander (2009) find similar results in 17 Latin American countries. Zetterberg (2009) focuses directly on quotas, finding no indications that political participation is higher in Latin American countries that have quotas as compared to countries that do not. Barnes and Burchard (2012) argue that this null finding of Zetterberg may be due to a weak link between quotas and increases in female representation. Hence, a basic and important first step in the present study will be to establish the effects of quotas on numerical representation, in order to put further findings into context. 2.2
Quotas in Latin America and Outcomes of Interest
Latin America is the region in the world where electoral quotas have been implemented most extensively. In 1991 Argentina became the first democratic country in the world to introduce a national electoral quota, and many other countries did the same in the late 1990s. Most of these introduced their first quotas in 1996 or 1997, and the spread of quotas in the region indicates a strong contagion effect (Escobar-Lemmon and Robinson 2005).1 2.2.1 What Determined Quota Introductions in Latin America? Investigating the effects of quotas raises the question of what determined the timings of quota introductions in Latin America. This is important if one wishes to consider to what extent the quota introductions were endogenous to changing gender equity norms in society. Previous authors have argued that recommendations from international organizations, such as the UN, the EU, the Inter-Parliamentary Union (IPU), and the OSCE have been highly influential in the introduction of quotas in Latin America (e.g., Dahlerup 2003), and so have international agreements, in particular the Platform for Action in 1995 at the fourth Women’s World Conference in Beijing (Htun and Jones 2001; Schwindt-Bayer 2009; True and Mintrom 1
Table A1 in the appendix shows the years of first quota introduction in Latin American countries.
9
2001). This platform urged governments to ensure equal access and full participation in political decision making, and the introduction of gender quotas was suggested as a specific measure to consider. In fact, the only countries to pass laws on quotas before the Beijing conference were Argentina and Bolivia. But why, then, have only some Latin American countries introduced quotas? To answer this question, it is insightful to look at the countries that currently do not have any quotas. For instance, gender quotas have been deemed unconstitutional in some countries. This happened in both Uruguay in 1988 and in Colombia in 2000, and in Venezuela the quota law introduced in 1997 was declared unconstitutional in 2000 and subsequently removed (quotaproject.org). However, Uruguay has now passed a new electoral law stipulating a 33 percent quota that will go into effect in the 2014 elections. In Chile, the former president Michelle Bachelet proposed a gender quota bill that did not pass in the parliament, and in Guatemala a quota bill did not reach the required two-thirds majority needed for implementation. Moreover, Nicaragua has a law on “the promotion of the necessary measures, in conformity with the Law of the matter, in order to establish a proportional percentage of women and men to the positions on the electoral lists of the national /…/ Parliament” (quotaproject.org), but no binding quota law. In El Salvador, women’s movements have lobbied for a quota law but without success. Hence, in several countries the reasons for not having introduced a quota are judicial, while others have come rather close to doing so. This, together with the fact that the quotas introduced have been of similar types, have been introduced at similar times, and the relatively similar historical backgrounds of the Latin American countries, lend some support to the idea of the pool of Latin American non-quota countries being a potential reference group for the quota countries. While the introductions of gender quotas in Latin America have clearly not been completely random, looking at the Latin American experience still provides a unique opportunity to study the effects of legal gender quotas in national elections. We discuss this
10
more extensively in the empirical strategy section below, and in the Appendix we also propose some ways to control for endogeneity of quota introductions by presenting some quantitative tests of their exogeneity and propose different flexible estimation strategies to be used. 2.2.2 What Outcomes Can We Expect Quotas to Affect? There is a lack of agreement in the literature with respect to which outcome variables are most likely affected by gender quotas (Wängnerud 2009) and also in in what direction they should be affected (Franceschet et al. (2012a). Hence, we will focus on a wide set of outcomes in this study, and in order to identify key areas that we think may have been affected by the quotas we will rely on both theory and previous empirical literature. Below, we motivate the two main areas covered: government policies and women’s political participation. 2.2.2.1 Government Policy The previous literature on gender and politics is full of suggestions about the effects of women in politics on government policy. For instance, Swiss et al. (2012) find that women’s representation is correlated with better child health as well as with increased vaccinations. Besley and Case (2003) report that male and female politicians in the US behave differently with regard to spending priorities, with women putting more focus on education and support for families and children. Similarly, Clots-Figueras (2011) studies elections in India where either a man or a woman closely beat an opponent of the other sex, and similarly finds that women invest more in education. Svaleryd (2009) shows that more female politicians in Swedish municipalities increase childcare expenditure relative to spending on the elderly. Although not dealing with female politicians per se, studies from the US have also found that granting suffrage to women increased both the size of the state in general (Lott and Kenny 1999) and health spending in particular (Miller 2008). As mentioned earlier, Chattopadhay and Duflo (2004) find that policies on issues closer to the preferences of women were
11
implemented more often in Indian villages with female chiefs (randomly allocated by a quota). However, when investigating the effects of gender quotas it is important to contextualize the expected effects, as gender roles are social products that differ widely across the globe and over time (Wängnerud 2009; Franceschet et al . 2012b). For instance, in the OECD countries, women had more conservative political preferences than men and tended to vote for bourgeois parties in the 1960s (Campbell et al. 1960). Since then, however, women’s political preferences have gradually shifted leftwards all over the OECD area, and in Scandinavia women are now more left-leaning than men (Inglehart and Norris 2000). A range of studies have revealed that gender gaps in political preferences are prevalent across the world and regarding a broad range of policies (e.g., Svallfors 1997; Alesina and La Ferrara 2005; Alvarez and McCaffery 2003; Lott and Kenny 1999; Aidt and Dallal 2008; Chattopadhyay and Duflo 2004), but they also show differences across countries and time periods. In Latin America – the region under study – women are more likely than men to vote for conservative parties (Escobar-Lemmon and Taylor-Robinson 2005). However, some differences between men and women are universal in direction (although not in degree), such as women conducting more paid and unpaid care work than men and also bearing greater responsibility for childcare. In line with this, Franceschet and Piscopo (2008) find that the gender quota introduction in Argentina seems to have induced an increase in parliamentary bills concerning women’s issues such as reproductive health and violence against women. On the other hand, they also find that this increased attention did not carry over to changes in policy outcomes. Taylor-Robinson and Heath (2003) do not find that women have a higher priority than men on children and family issues in Honduras, but that they more frequently debate women’s rights than men do. Hence, there is good reason to believe the effects of quotas to be context dependent.
12
As we expect that policies on issues closer to the preferences of women are more likely to be affected, it is important to get a picture of preference differences between men and women in Latin America. This also makes the analysis closer to the definition of substantive representation, implying a policy responsiveness to women’s interests (e.g. Weldon 2002). Looking at data from the Latinobarometer survey (described in the data section), women are more inclined than men to think that the state is responsible for health care and for education, but they also perceive that taxes are too high to a greater extent. 2 Hence, while spending priorities may very well differ depending on women’s representation, we should not expect the results from other contexts to automatically carry over to Latin America. 2.2.2.2 Women’s Political Participation, Confidence, and Satisfaction. One part of the motivation of gender quotas in politics is that female politicians may act as role models and pave the way for other women in politics. Empirical evidence from India supports this view: Beaman et al. (2009) show that people’s gender stereotypes are weakened and that attitudes toward women as policymakers become less biased once they have been exposed to female leaders following a quota at the municipal level, supplying robust evidence that quotas may be effectively used to promote gender equality by reducing the gender bias of societal norms. An important question is whether this result can be generalized to other contexts and to national level quotas. Evidence from a wide range of contexts show large differences in popular political participation between the sexes (e.g. Burns et al., 2001; Inglehart and Norris, 2003; Barnes and Burchard, 2012; Isaksson et al., 2014). Consulting the Latinobarometer data once more, we see clear differences between men and women in political participation also in Latin America prior to the quotas. 3 For instance, women were less likely to participate in demonstrations and to vote, and they were more likely to state that politics is complicated, 2 3
These results are available upon request. Again, the results are available upon request.
13
that they were not interested in politics, and that they did not engage in politics in general. Given these apparent inequalities, it is important to investigate whether an increased share of women in parliaments following quotas has affected women’s political interests and/or ability to participate in the political sphere. Using the Latinobarometer data from 2005, Zetterberg (2009) shows that political participation is not higher in countries that have introduced quotas. While he also controls for lagged (from 1996) control variables, he does not fully exploit the data to include all years possible. Furthermore, he does not make use of country fixed effects or control for trends in the variables of interest. As will be explained in the empirical strategy, we will employ difference in differences estimations (and also triple differences by exploiting men as another control group), and hence we believe that there may be substantial benefits from updating these findings. 3 3.1
Data and Empirical Strategy Data
The dependent variables are intended to cover our two main areas of investigation, namely policies and political participation. These variables, as well as our independent variables of main interest discussed below.4 Using data from the Global Database of Quotas for Women (quotaproject.org) and the country-specific references therein, we classify a country as a quota country if it introduced some legal quota for women in its national parliament during our period of investigation, i.e., 1985-2009. We also create two variables representing the size of the quotas in the lower (or only) house (Quota size lower) and upper house of parliament (Quota size upper), measuring the share of candidates that are required to be women. They range from 0 for periods and countries with no quotas to 1 in the hypothetical case of a quota that requires all candidates to be women. The data on the share of women in parliament is taken from the PARLINE 4
Table A2 in the appendix gives an overview of the variables.
14
database on national parliaments from the Inter-Parliamentary Union (www.ipu.org). Again, we separate this measure into women in the lower house (WiP lower) and upper house (WiP upper) of parliament. The data on quotas and women in parliament is available for all years of investigation. We also use data on the share of women ministers (female ministers) as used by Escobar-Lemmon and Taylor-Robinson (2005). This data includes all our countries for all years of democracy from 1980 to 2003. To operationalize the effects of quotas on policies, we first look at government consumption as a share of GDP. This measure, as well as the total tax revenues, should be a good proxy for the size of the state. We also investigate effects on spending on certain sectors of the economy, namely on education, health, social security, and social spending in general (which is the sum of the former three plus housing expenses). For these spending measures we have data since at least 1990 from the World Development Indicators (WDI) and the Economic Commission for Latin America and the Caribbean (ECLAC). To investigate the effects of quotas on political participation, we mainly use data from the Latinobarometer. The Latinobarometer is an annual survey (with a gap in 1999) that started in 1995 and now includes 18 countries in Latin America, with about 20,000 respondents per wave. In 1995 only eight countries were included in the survey, but already in 1996 the number was increased to seventeen.5 In total, we have 216,998 observations from 18 countries from the years 1995 to 2007. A disadvantage of this survey material is that for the group of countries that passed quota laws around 1997, there is only data for a few years prior to the first quota election. The variables we focus on in assessing the effects on political participation are frequency of talking about political issues with friends, political interest, and whether the respondent would vote if there were an election the Sunday after the interview. We also look at confidence in the national congress and satisfaction with democracy, which 5
The only country under study not included in 1996 is the Dominican Republic, which was included for the first time in 2004.
15
are thought to reflect to what extent men and women are satisfied with the democratic system and its institutions. For all variables from the Latinobarometer, we use the (weighted) average for women in each country and year. Using micro-level data, it is also possible to create measures on the gender gap in participation, which in turn enables us to use the “difference in difference in differences” approach further elaborated on in the empirical strategy section below. All gap measures are created as the average value for men less the average value for women in the same year and country. As a complementary measure on political participation, we also include official figures on electoral turnout from the IDEA voter turnout database. Unfortunately, this data is not measured by gender, so we can only observe the aggregate development of men and women before and after quotas. 3.2
Empirical Strategy
In order to investigate the effects of quotas on the outcomes described above, and to assess the actual mechanism of increased women in parliaments, we proceed in several steps. Since a vital part of this study relies on quotas being effective in raising the number of women in national parliaments, we will start by discussing how to assess this assumption. We then move on to describe our strategy for estimating the effects of gender quotas on other variables. 3.2.1 Quotas and female representation in Latin America Though investigated in previous literature, it is not evident to what precise extent gender quotas in general have actually increased the number of women in parliaments in Latin America. Whereas some countries in other parts of the world have introduced gender quotas in the form of reserved seats guaranteeing women a specific share of the seats in parliament, the quota countries in Latin America have all introduced so-called list quotas where a specified share of all party candidates must be women. This makes the actual outcome in terms of representation uncertain. Moreover, the countries within the region also differ in terms of whether voters are free to choose which candidates on a party’s list to vote for or
16
whether the orders of candidates are fixed (open vs. closed lists), whether there is a placement mandate implying that parties have to place women on certain (more electable) positions on the lists, and to what degree there are sanctions against parties that do not comply with the quota law. Jones (2009), looking at data at the election district level in Latin America, shows that in Latin America the effects of national election quotas on representation vary depending on both quota and election rules, as well as district size. However, the main conclusion is that as long as there is enforcement, both open and closed list quotas are effective, even if closed lists are found to be slightly more so. As dividing quotas into categories may create almost as many categories as there are elections at the national level, we keep the analysis simple by defining only the size of the legal quota irrespective of the rules surrounding it. In order to see to what extent quotas have been effective, we start by estimating the effect of quotas on the share of women in national parliaments, which will serve several purposes: firstly, it will help us understand to what extent quotas have been effective in Latin America and hence guide us on how to interpret later results on quota effects; secondly, it will give us an empirical understanding of what the processes of quota implementations looked like; and thirdly, it will serve as a benchmark test of whether significant effects of quotas can be identified within our sample. Our benchmark estimations for the share of women in parliament for country c at time t take the forms: WiPct c t 1 Qct ct ,
(1)
where WiP is the share of women in parliament in country c in year t, c are country fixed effects, t are year fixed effects, Q is one of our quota measures, where we use either Quota ineffect which is a dummy variable equal to one when a country has a quota in effect, or Quota size which measures the size of a country’s quota in each time period (i.e. it equals zero in years when a country has no quota in effect, and is 0.30 for a country in a year when it has
17
a 30 percent quota in effect). This is a typical “difference-in-differences” (DiD) setup, where the assumption is that all countries would follow the same path (albeit from different starting points) in the absence of quotas. In the appendix section A2 we discuss and present results from a more flexible specification using country-specific time trends ( c t ) . In the Appendix section A1 we also propose two tests of the endogeneity of the quota introductions to establish to what extent the quotas can be seen as exogenous to other processes related to women’s representation. In particular, we show in Table A3 that there are no pre-quota differences in the number of women in parliament. Even though this test is not a necessary condition for our empirical strategy, as we control for level differences between quota and non-quota countries, it further increases our confidence that the quota introductions are not merely picking up pre-existing trends. We also show in a placebo regression that the introduction of quotas in the lower houses has no measurable impact on WiP in the upper house. This further increases our confidence that the results are not driven by omitted variables, such as simultaneous changes in voter or party attitudes, since such changes should plausibly also affect the number of women in the upper house.
3.2.2 The Effects of Quotas on Policy and Political Participation Turning to our estimations of the effects of quotas on policy and political participation, we again start by estimating DiD models. Hence, for any dependent variable y ct , we estimate the following models: yct c t 1 Qct ct ,
(2)
where we start with a standard DiD setup with year and country fixed effects. As discussed before, this setup is robust to common non-linearities in the trend and assumes the countries to have similar developments in the absence of quotas. It should also be noted that the
18
introduction of country fixed effects here constitutes a major improvement over cross country studies by allowing for pre-quota differences in levels of the outcome variables. 6 For the variables based on micro-level data from the Latinobarometer (Notvoting, Donttalkpolitics, Notinterestedpol, Noconf.congress, and Satisfieddemocracy), we create gap measures (the average of males minus the average of females) in order to exploit yet another difference, thereby creating DDD (difference-in-difference-in-differences) estimators. In these cases the estimates of 1 will show us whether the quota caused women’s development to differ from that of men.7 4 4.1
Results Quota Effects on the Share of Women in Parliament
Table 1 presents results from regressions to establish the size of the effect of introducing women quotas on the actual election outcomes in terms of the share of women in parliament. Column 1 shows the results from the difference in differences estimation of the share of women in parliament (henceforth WiP) on a quota dummy and column 2 presents the corresponding results using the size of the quota. We see in Column 1 that the effect of introducing quotas for women in the lower (or only) house of parliament is on average 7.2 percent more women, and Column 2 shows that imposing a quota on average leads to an increase of around 25 percent of the quota size. Columns 3 and 4 show a regression of WIP on a time trend common to all countries, and a quota dummy and the size of the quota, respectively. We see that, on average, there is a small yet positive and statistically significant time trend, implying that the share of women in 6
In the appendix we also conduct a robustness check by allowing for both year fixed effects and country-specific time trends, t c t . 7
Bertrand et al. (2004) caution against trusting estimates from DiD regressions in finite samples if there is reason to believe that there is serial correlation in variables and error terms. Hence, we use the Newey-West estimator for standard errors, assuming heteroskedastic error terms and serial correlation up to two periods back. As a robustness check, we also clustered the standard errors at the country level (results are available upon request), yet the interpretations of our results remained unchanged.
19
parliaments increases by about 0.5 percentage points per annum in the absence of gender quotas. Imposing a gender quota, however, yields an increase of 0.22 of the value of that quota (Column 4). That is, a quota stipulating that at least 30 percent of the candidates in an election ought to be women raises the share of women in parliament by about 7 percentage points on average. At a first glance, this may seem like a rather small figure. However, one should remember at least three things: that countries already had some women in parliament prior to the quota introductions; that in many cases there is no placement mandate of candidates; and that voters may often choose whom to vote for among the candidates on the lists. Moreover, if one compares the effect of the quota to the time trend, the result of a 30 percent quota is equivalent to an instant increase of the share of women in parliament that would have taken about fifteen years to achieve without it.8 4.2
Quota Effects on Policy and Political Participation and Potential Mechanisms
Turning to our results from the reduced form regressions of policy and political participation, Table 2 shows the difference in differences results of regressing our outcome variables on a quota dummy (Column 1) or on the size of quotas in the lower (or only) house of parliament (Column 2). As is evident, we find very few significant results here. In fact, the only coefficient that is statistically significant with both types of measures is the one for Government consumption as a share of GDP. The results for this coefficient suggest that introducing a quota for women in parliament is associated with an increase of almost two percentage points in government consumption. In the Appendix (Table A5) we show results when adding country specific time trends and we note that Government consumption – the only consistently significant outcome from the DiD setup – loses its statistical significance
8
To save space, the results from identical estimations on the effects of quotas in the upper house are not shown, yet are available upon request. The results are almost identical to those for the lower house.
20
once we control for country specific time trends. Hence, the results indicate that there seem to have been few clear and homogenous effects across the Latin American countries. So far we have found that quotas are effective in raising the number of women in parliaments, but that they do not seem to have had any clear statistically significant effects on policy, attitudes, or political participation. As a first step in assessing why this is so, we will test whether the share of women in Latin American parliaments in general are correlated with our outcome variables, thus investigating whether
the relationship between female
representation and our outcomes differ depending on whether or not women have been elected with or without the help of quotas. Any findings of correlations between female representation and our policy and attitudes variables would be consistent with either quota women being different from non-quota women in at least some of the countries included, that they have become marginalized by the incumbent male elite, or that the correlation between the non-quota share of women in parliament and the outcomes investigated are spurious. These explanations are, of course, not mutually exclusive. We perform estimations as similar as possible to those already discussed in Table 2, by regressing our outcome variables on the share of women in parliament directly, using observations only from years in which there was no legal quota in effect.9 In doing this, we will identify the correlation between our outcome variables and the share of women in parliament that stems from the ordinary election process. The results of these estimations are presented in Column 3 of table 2.10 As seen, quite a few of the parameters become significant when using the share of women in parliaments pointing to correlations between higher female representation and policy, increases in women’s political interest and participation, and increased confidence in congress in our
9
This is essential – if we were to include quota years, the quota effects would make up the lion part of the variation in the share of women in parliament. 10 The corresponding results when adding country specific time trends are shown in column 3 of Appendix Table A5.
21
sample of countries. These results are interesting in themselves, as the estimations do control for country fixed effects, thereby mitigating some of the endogeneity concerns of the share of women in parliament. More importantly for our purposes, however, is the fact that these same correlations can not be found when looking at the introduction of legal quotas for women’s representation, which seems to indicate that that female representation that stems from gender quotas is somehow different from that of an ordinary election process. One potential explanation for why quota representation is different from ordinary representation is, as discussed in the theoretical section, that the incumbent elite reacts to quotas by marginalizing women (Childs and Krook 2006; Dahlerup 2006; Grey 2006; Zetterberg 2008). In order to empirically test whether the quota introduction has led to increased political power for women, we investigate the effects of the quotas on the share of women in ministerial positions. Arguably, if women have increased their real political power this should show up also at the ministerial level, which in itself is unaffected by the quotas. Using the data collected by Escobar-Lemmon and Robinson (2005), which contains the share of women in ministerial positions from 1980, i.e., at a point before any country had introduced quotas, to 2003, when almost all quota countries had their quotas in effect, we test the effects of quotas on the share of female ministers in the same way as we have already done when looking at the effects of quotas on the share of women in parliament. Table A6 in the Appendix shows that there are no statistically significant effects of quotas on the share of women in ministerial positions in any of the specifications tried. Hence, it seems that a possible mechanism for the insignificant results for our outcome variables is that the elected women are marginalized from power and hence have little influence on policies. To sum up, our evidence of no increased shares of women ministers seems to point at the women elected with the help of quotas being marginalized and holding relatively little political power. However, there may also be other differences between women elected via
22
quotas and women elected via regular electoral competition as discussed in section 2.1. Moreover, it is also possible that the correlations between the share of women in parliament and the outcome variables are spurious in that higher female representation and other outcomes are all driven by unobserved factors that are not correlated with the introduction of gender quotas. Whether quota women are marginalized or are different from women elected through an ordinary election process, our results tell us that using gender quotas as a “fast track” to gender-equal representation may not – at least over the time periods spanned here – be very successful other than in terms of nominal representation. If, on the other hand, it is the case that the correlations between the share of women in parliament and policy are largely spurious, it seems that female representation – while likely desirable for other reasons – may not be very important for actual policy, women’s political interest and participation. The findings thus point to the complexity and uncertainty that still exist around the important yet sensitive issue of women’s quotas in elections. 5
Conclusions
Previous research has shown that quotas for women in politics affect the distribution of public goods, attitudes toward women, and the probability of women being elected even after the quotas are withdrawn. However, with respect to the present study these findings are based on other types of quotas, for other levels of government and mostly in other countries. In this article we assess the effects of quotas in Latin American national parliaments on a wide range of outcomes. Although we find substantial and statistically significant effects of quotas on the number of women in parliament, we find very few statistically significant results on political participation and policies. Furthermore, all the effects of quotas, apart from the direct effect on the share of women in parliament, are sensitive to model specification. The lack of
23
significant effects indicates that the increased share of women in national parliaments resulting from quotas has had few measurable effects beyond numerical representation in Latin America. This conclusion is also strengthened by the fact that when using the share of women in parliament directly, we find correlations with policies and political participation more in line with expectations, indicating that to the extent that these correlations are causal effects, they do not seem to carry over to an increased share of women in parliament following a quota. It is of course possible that the quotas have had effects on indicators that we do not measure. In particular, it is common to measure substantial representation as proposing bills, co-sponsoring bills, or participate in debates linked to women’s interests such as e.g. parental leave or domestic violence (e.g. Weldon 2002). Conducting such an analysis is difficult for a large sample of countries and we argue that if quotas had significant effects on these variables, they should also affect spending. That we find correlations with non-quota female politicians and spending reinforces this point, and also the fact that non-quota female participation is likely to have co-evolved with other societal factors. In particular, Weldon (2002) finds that women’s movements are more effective in affecting legislation than is numerical representation. Our finding contrasts some of the previous studies in the field.
For instance,
Chattopadhay and Duflo (2004) and Beaman et al. (2009) found that quotas at the lowest level of governance (village councils) in India affected the distribution of public goods in favor of women’s preferences and also reduced negative stereotypes toward women. Our findings also highlight the fact that the results of similar rules may be different in different contexts, and looking at the preferences of women in Latin America, it is far from evident that one should expect a larger government or more spending on health or education as has been proposed before. Nonetheless, not taking quota introductions into account, we are often able to reproduce statistically significant correlations of women in parliament and our outcomes,
24
which may indicate either that the correlations are spuriously driven by other time-variant factors (we do control for country fixed effects), or that increasing women in parliaments via quotas has some other adverse effects. For instance, it is possible that at the national level, women who are placed on the party lists following a quota are those closest to male-leaning party elites. Hence, quotas in national elections may not be effective in increasing the effective representation of women’s preferences, but only in increasing the number of women in parliament. Finally, the types of quotas studied here are also different from those in the previous literature. The quotas in Latin America say nothing about the role of the women once they have been elected, which is in strong contrast to for instance the quotas at the Indian village level which allocate chief positions – including agenda setting power – to women. Our result that the quotas in Latin America did not lead to increased power for women in terms of ministerial positions indicate that the agenda setting power of quota women may indeed be low. To sum up, the evidence presented here does not allow us to establish with certainty whether the lack of significant effects are due to marginalization of women, due to women elected with and without quotas being different from each other, or to simple correlations between the shares of (non-quota) women in parliament and our outcomes being spurious. It is also possible that the effects of gender quotas take a longer time than investigated here to materialize. For instance, in a study on the UK, Childs and Krook (2012) have found that the stigma of being a quota woman indeed lessened over time. Either way they do indicate that using gender quotas as a “fast track” to gender equal representation has not been successful other than in terms of numerical representation in Latin America. Based on existing levels of knowledge, it is too early to draw conclusions about the importance of agenda-setting power,
25
level of governance, or context more generally, and we therefore urge future studies to further investigate the causal effects of different types of quotas in other settings.
26
References Aidt, T. S. and Dallal, B. 2008. “Female Voting Power: The Contribution of Women’s Suffrage to the Growth of Social Spending in Western Europe (1869-1969)” Public Choice, Vol. 134, No. 3-4, pp. 391-417. Alesina, A. and La Ferrara, E. 2005. “Preferences for Redistribution in the Land of Opportunities.” Journal of Public Economics, Vol. 89, No. 5-6, pp. 897-931. Alvarez, R. M. and McCaffery, E. J. 2003. “Are There Sex Differences in Fiscal Political Preferences?” Political Research Quarterly, Vol. 56, No. 1, pp. 5-17. Araújo, C. and García, A. 2006. ”Latin America: The experience and impact of quotas in Latin America”. Chapter 4 in Dahlerup, D. (ed) 2006. ”Women, Quotas, and Politics” Routledge, New York. Barnes, T. D. (2012). “Gender and Legislative Preferences: Evidence from the Argentine Provinces.” Politics & Gender, 8(04), 483-507.Barnes, T. D., & Burchard, S. M. (2013). “’Engendering’ Politics The Impact of Descriptive Representation on Women’s Political Engagement in Sub-Saharan Africa.” Comparative Political Studies, 46(7), 767-790. Beaman, L., Chattopadhyay, R., Duflo, E., Pande, R. and Topalova, P. 2009. ”Powerful Women: Does Exposure Reduce Bias?” The Quarterly Journal of Economics, Vol. 124, No. 4, pp. 1497-1540. Bertrand, M., Duflo, E. and Mullainathan, S. 2004. “How Much Should We Trust Differences-in-Differences Estimates?” The Quarterly Journal of Economics, Vol. 119, No. 1, pp. 249-275. Besley, T., and Case, A. 2003. ”Political Institutions and Policy Choices: Evidence from the United States”. Journal of Economic Literature, Vol. 41, No. 1, pp. 7-73. Besley, T., and Coate, S. 1997. ”An Economic Model of Representative Democracy” The Quarterly Journal of Economics. Vol. 112, No. 1, pp.- 85-114. Besley, T., Pande, R. , Rahman, L. and Rao, V. 2004. ”The politics of public goods provision: Evidence from Indian local government.” Journal of the European Economic Association. Vol. 2, No. 2-3, pp. 416-426. Burns, N., Schlozman, K. L., & Verba, S. (2001). The private roots of public action: Gender, equality, and political participation. Cambridge, MA: Harvard University Press. Campbell, A., Converse, P. E., Miller, W. E. and Stokes, D. E. 1960. The American Voter. New York: Wiley. Chattopadhyay, R. and Duflo, E. 2004. “Women as Policy Makers: Evidence from a Randomized Policy Experiment in India” Econometrica, Vol. 72, No. 5, pp. 1409–1443. Childs, S., and Krook, M-L. 2006. ”Should Feminists Give Up on Critical Mass? A Contingent Yes” Politics & Gender, Vol. 2, No. 4, pp. 522-530. Childs, S., and Krook, M. L. (2009). Analysing women's substantive representation: From critical mass to critical actors. Government and Opposition, 44(2), 125-145. Clots-Figueras, I. 2011. “Women in politics: Evidence from the Indian States”, Journal of Public Economics, Vol. 97, No 7-8, pp. 664-690.
27
Dahlerup, D. 2006. ”The Story of the Theory of Critical Mass”. Politics & Gender, Vol. 2, No. 4, pp. 511-522. Del Campo, E. (2005). Women and politics in latin america: Perspectives and limits of the institutional aspects of women's political representation. Social Forces, 83(4), 1697-1725. Desposato, S., & Norrander, B. (2009). “The gender gap in Latin America: Contextual and individual influences on gender and political participation.” British Journal of Political Science, 39, 141-162. Diamond, I., & Hartsock, N. (1981). “Beyond Interests in Politics: A Comment on Virginia Sapiro's’ When Are Interests Interesting? The Problem of Political Representation of Women’.” The American Political Science Review, 75(3), 717-721. Downs, A. 1957. An Economic Theory of Democracy, New York: Harper Collins Edlund, L. and Pande, R. 2002. “Why have Women Become Left-Wing? The Political Gender Gap and the Decline in Marriage” Quarterly Journal of Economics, Vol. 117, No. 3, pp. 917-961. Edlund, L., Haider, L. and Pande, R. 2005. “Unmarried Parenthood and Redistributive Politics.” Journal of the European Economic Association, Vol. 3, No. 1, pp. 95-119. Escobar-Lemmon, M. and Taylor-Robinson, M. 2005. “Women Ministers in Latin American Government: When, Where, and Why?” American Journal of political Science, Vol. 49, No. 4, pp. 829-844. Fallon, K. M., Swiss, L., & Viterna, J. (2012). Resolving the Democracy Paradox Democratization and Women’s Legislative Representation in Developing Nations, 1975 to 2009. American Sociological Review, 77(3), 380-408. Finseraas, H., Jakobsson, N., and Kotsadam, A. 2012. “The Gender Gap in Political Preferences: An Empirical Test of a Political Economy Explanation”. Social Politics 19(2): 219-242. Fréchette, G., Maniquet, F., and Morelli, M. 2008. ”Incumbents’ Interests and Gender Quotas”. American Journal of Political Science, Vol. 52, No. 4, pp. 891-907. Franceschet, S., and Piscopo, J. 2008. “Gender Quotas and Women’s Substantive Representation: Lessons from Argentina” Politics & Gender, No. 4, pp. 393-425. Grey, S. 2006. ”Numbers and Beyond: The Relevance of Critical Mass in Gender Research”. Politics & Gender, Vol. 2, No. 4, pp. 492-502. Heath, R. Schwindt-Bayer, L., and Taylor-Robinson, M. 2005. ”Women on the Sidelines: Women’s Representation on Committees in Latin American Legislatures”. American Journal of Political Science, Vol. 49, No. 2, pp. 420-436. Htun, M and Jones, M. 2001. "Engendering the Right to Participate in Decision-making: Electoral Quotas and Women's Leadership in Latin America." In Craske, N. and Molyneux, M. (ed) “Gender and the Politics of Rights and Democracy in Latin America” Palgrave, London. Inglehart, R. and Norris, P. 2000. “The developmental theory of the gender gap: Women's and men's voting behavior in global perspective.” International Political Science Review, Vol. 21, No. 4, pp. 441-463. Inglehart, R., & Norris, P. 2003. Rising tide: Gender equality and cultural change around the world. Cambridge, UK: Cambridge University Press.
28
Isaksson, A. S., Kotsadam, A., & Nerman, M. 2014. “The Gender Gap in African Political Participation: Testing Theories of Individual and Contextual Determinants.” Journal of Development Studies, Volume 50(2). Jones, Mark P. (1999). Assessing the effectiveness of gender quotas in open-list proportional representation electoral systems. Social Science Quarterly, 80/2, 341– 355. Jones, M. P. (2009). Gender quotas, electoral laws, and the election of women evidence from the Latin American vanguard. Comparative Political Studies, 42(1), 56-81. Kittilson, M. C. (2008). Representing women: The adoption of family leave in comparative perspective. The Journal of Politics, 70(02), 323-334. Krook, M. L. 2006. “Reforming representation: The diffusion of candidate gender quotas worldwide.” Politics & Gender, 2(3), 303-327. Lott, J. and Kenny, L. 1999. ”Did Women's Suffrage Change the Size and Scope of Government?” The Journal of Political Economy. Vol. 107, No. 6, pp. 1163-1198. Miller, G. 2008. “Women's Suffrage, Political Responsiveness, and Child Survival in American History” The Quarterly Journal of Economics, Vol. 123, No. 3, pp. 12871327. Pande, R. 2003. “Can mandated political representation increase policy influence for disadvantaged minorities? Theory and evidence from India.” American Economic Review, Vol. 93, No. 4, pp. 1132-1151. Phillips, Anne. 1995. The Politics of Presence. Oxford: Clarendon Press. Piscopo, Jennifer M. 2011. “Rethinking Descriptive Representation: Rendering Women in Legislative Debates.” Parliamentary Affairs 11 (3): 1–25. Sapiro, V. (1981). “Research Frontier Essay: When Are Interests Interesting? The Problem of Political Representation of Women”. The American Political Science Review, 701-716. Schwindt-Bayer, L. A. 2009. ”Making Quotas Work: The Effect of Gender Quota Laws On the Election of Women” Legislative Studies Quarterly, Vol. 34, No. 1, pp. 5-28. Svaleryd, H. (2009). “Women's representation and public spending.” European Journal of Political Economy, 25(2), 186-198. Svallfors, S. 1997. “Worlds of Welfare and Attitudes to Redistribution: A Comparison of Eight Western Nations.” European Sociological Review, Vol. 13, No. 3, pp. 283-304. Swiss, L., Fallon, K. M., & Burgos, G. (2012). Does Critical Mass Matter? Women's Political Representation and Child Health in Developing Countries. Social forces, 91(2), 531-558. Taylor-Robinson, M. M., & Heath, R. M. (2003). Do women legislators have different policy priorities than their male colleagues? A critical case test. Women & Politics, 24(4), 77101. Thomas, S. 1991 “The Impact of Women on State Legislative Policies” The Journal of Politics, Vol. 53, No. 4, pp. 958-976. Tripp, A. M., & Kang, A. (2008). “The Global Impact of Quotas On the Fast Track to Increased Female Legislative Representation.” Comparative Political Studies, 41(3), 338-361. True J and Mintrom M. 2001. “Transnational Networks and Policy Diffusion: The Case of Gender Mainstreaming”. International Studies Quarterly. Vol. 45, No. 1, pp. 27-57. Weldon, S. L. (2002). Beyond bodies: Institutional sources of representation for women in
29
democratic policymaking. The Journal of Politics, 64(04), 1153-1174. Wängnerud, L. 2009. ”Women in Parliaments: Descriptive and Substantive Representation”. Annual Review of Political Science, Vol. 12 ,pp. 51-69. Zetterberg, P. 2008. ”The Downside of Gender Quotas? Institutional Constraints on Women in Mexican State Legislature”. Parliamentary Affairs, Vol. 61. No. 3, pp. 442-460. Zetterberg, P. 2009. ”Do Gender Quotas Foster Women’s Political Engagement?”. Political Research Quarterly, Vol. 64, No. 4, pp. 715-730.
30
Tables Table 1. WiP lower. Fixed effects estimations of the share of women in the lower (or only) house of parliament. (1) (2) (3) (4) Quota ineffect lower 0.072*** 0.058*** (0.021) (0.019) Quota size lower 0.249*** 0.216*** (0.067) (0.048) year dummies yes yes time
0.005*** 0.005*** (0.001) (0.001) Constant 0.061*** 0.061*** 0.049*** 0.050*** (0.015) (0.015) (0.013) (0.013) Observations 115 115 115 115 R-squared 0.729 0.741 0.623 0.641 Number of countries 18 18 18 18 Robust standard errors in parentheses. *** p
